Buy Or Wait Evidence Agent — Development skill for Claude Code

Buy or Wait Evidence Agent

A financial decision agent built for HackerRank Orchestrate (September 2026), using bounded Anthropic evidence extraction and a deterministic cash-flow engine.

See [setup and architecture](code/README.md) for installation, environment variables and the complete request walkthrough. From the repository root, run `python3 code/main.py final` to generate and validate `output.csv`.

Buy or Wait?

Build an AI-powered financial agent that decides whether a user can safely afford a requested expense.

A user may ask: **"Can I afford this laptop?"**

Answering well takes more than the current balance. The agent must account for recurring expenses, pending payments, essential spending, confirmed income, available payment options, and relevant details buried in messages and images.

For every request, the agent decides whether the user should pay in full, pay partially, use installments, wait, or not proceed. The recommendation must be personalized: two users with the same balance can deserve different answers based on their commitments, priorities, payment preferences, and willingness to adjust flexible expenses.

A recommendation is safe only if the user can complete the full payment plan, cover essential expenses, and stay above their preferred minimum balance throughout the forecast period.

Read [`problem_statement.md`](./problem_statement.md) for the full task spec, input/output schema, allowed values, conflict-resolution rules, and submission format.

---

Quick Start

Clone the repository and move into the project directory:

git clone https://github.com/interviewstreet/hackerrank-orchestrate-september26.git
cd hackerrank-orchestrate-september26

Build your solution in `code/main.py`, or use another language and document its entry point clearly.

Your solution must:

• Read the input files from dataset/
• Generate one prediction for every request
• Write the final predictions to output.csv in the repository root

Ru